Sign In

Communications of the ACM

ACM TechNews

Novel Deep Learning Framework for Symbolic Regression

View as: Print Mobile App Share:
Representation of a a new deep reinforcement learning framework for a type of discrete optimization called symbolic regression.

Lawrence Livermore National Laboratory computer scientists have developed a new framework and accompanying visualization tool that leverages deep reinforcement learning for symbolic regression problems.


Computer scientists at Lawrence Livermore National Laboratory (LLNL) have created a new framework and visualization tool that applies deep reinforcement learning to symbolic regression problems.

Symbolic regression, a type of discrete optimization that seeks to determine the underlying equations or dynamics of a physical process, generally is approached in machine learning and artificial intelligence with evolutionary algorithms, which LLNL's Brenden Petersen said do not scale well.

LLNL's Mikel Landajuela explained, "At the core of our approach is a neural network that is learning the landscape of discrete objects; it holds a memory of the process and builds an understanding of how these objects are distributed in this massive space to determine a good direction to follow."

The team's algorithm outperformed several common benchmarks when tested on a set of symbolic regression problems.

From Lawrence Livermore National Laboratory
View Full Article


Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account